Social network analysis and educational change: unravelling the role of innovative teaching staff in a higher education environment

ABSTRACT

As the higher education landscape has shifted and calls for change and improvements have grown louder, researchers have turned to empirical investigations of what makes change possible in higher education. Recent studies on the impact of informal interactions for promoting change have given impetus to social network analysis as an approach to studying higher education change. However, the relationship between teachers’ social network characteristics and their disposition to adopt changes remains largely unknown. In this study, we describe the characteristics of teaching networks and explain the network characteristics of the most innovative teaching staff. By capturing whole teaching and research networks within a STEM discipline department at a Norwegian university, we show that teachers tend to speak to the same colleagues about both research and teaching, and that teaching communities are constrained by departmental sub-units (i.e. research groups). Furthermore, by combining the network data with a survey on diffusion of innovation, we find that the most innovative teachers (1) are central in the network, having the largest personal networks, (2) have a substantial reach to the rest of the network, and (3) are highly interconnected. This paper contributes to the understanding of change theories in higher education and discusses how these results can inform the practice of change agents and educational developers. In addition to confirming the presence of a research-teaching nexus, we discuss why leveraging the networks of the most innovative teaching staff would be a promising strategy for facilitating change in higher education.

KEYWORDS:

• Higher education
• social network analysis
• diffusion of innovation
• teaching networks
• change

Introduction

Globalization, digitalization, and massification of higher education have vastly altered the landscape of teaching and learning. Calls for quality enhancement have resulted in numerous initiatives that seek to improve student learning outcomes by promoting changes in faculty teaching practices (European Higher Education Area 2015; European Students’ Union 2015). Although accurately estimating the success of change initiatives can be challenging, it's clear that many organizations face difficulties in effectively implementing change, as it seems that a greater number of these initiatives tend to fail rather than succeed (Burnes 2011; Kezar 2018).

In the growing attention for understanding change in higher education, researchers have also looked towards different theoretical perspectives that originated outside of the educational field (Brankovic and Cantwell 2022). One such perspective through which change can be studied is Social Network Analysis (SNA). The appeal of SNA can be attributed, in part, to its capacity for quantitatively studying relational ties among social entities (Wasserman and Faust 1994).

This capacity aligns with the crucial role of sensemaking in organizational change (Weick 1995), because social relationships are one avenue through which teachers regularly construct, maintain, or change their understanding (Roxå and Mårtensson 2009) and transformational change in higher education is often about meaning (re)construction (Kezar and Eckel 2002). While it is undeniable that teachers engage in informal conversations with their peers about teaching and research, it is important to acknowledge the value of empirically exploring this aspect of academia in greater depth. Roxå, Mårtensson, and Alveteg (2011) emphasize that to understand change processes within academic organisations, we must understand the importance of various influential factors, including the individuals involved in that process.

Earlier research focused on the use of social network theory to study change largely in pre-college educational settings (Daly et al. 2010; Daly and Finnigan 2010). More recently, there has been an increase of studies employing social network theory with focus on change in the context of higher education (e.g. Jippes et al. 2013; Knaub, Henderson, and Fisher 2018; Mladenovici et al. 2022; Van Waes et al. 2016). Change processes are often socially constructed and these social processes can facilitate or inhibit change efforts (Daly and Finnigan 2010). Therefore, the aim of our study is to investigate how teachers’ networks at a department level relate to their disposition towards adopting changes and to understand the relevance of these networks for promoting change. We have focused our attention at the department level, which has been identified as highly conducive to creating change (Quardokus and Henderson 2015), as it can provide teachers with a sense of ownership and relevance, and agency (Annala et al. 2023; Keesing-Styles, Nash, and Ayres 2014). Keeping the analytical focus on the department enables us to represent an entire bounded network (Clifton and Webster 2017), which allows us to analyze not only individual teachers’ personal networks, but also their relative position within a larger network.

We have also decided to focus on the informal teaching conversations because SNA studies on higher education change suggest that informal networks have a significantly stronger impact than formal organization on individuals’ decision to engage in change behavior (Kezar 2014) and challenge the underlying belief that the formal institutional system carries the majority of the impact (Daly and Finnigan 2010). Previous literature further shows that informal conversations can be of great significance for understanding teaching practice and culture (de Lima 2010; Roxå, Mårtensson, and Alveteg 2011; Thomson and Trigwell 2016). Despite the importance of analyzing social interactions in order to better understand change efforts (Daly et al. 2010), there is still scarce knowledge about the role of social structures for change processes specifically in higher education.

In our study, we will focus on the Diffusion of Innovation theory that describes how the adoption of an idea (or a product) spreads through a social system (Rogers 1983). Combining an explicit change theory with social network analysis could provide meaningful insight into network characteristics of teachers with different dispositions to adopt changes: how central or peripheral they tend to be, how interconnected they are, or how frequently they engage in the network. Therefore, by combining network and change theories, the purpose of this study is to investigate how SNA can help in understanding teachers’ disposition towards introducing changes in higher education. We aim to achieve that by providing answers to following research questions:

1. What are the characteristics of teachers’ informal teaching networks within a departmental network?

2. What is the relationship between teachers’ social networks and their disposition to adopt changes in teaching?

Our findings point towards a strong interplay between teaching and research networks and emphasize the role of formal organizational sub-units as the hubs for teaching conversations. We also demonstrate that most innovative teachers have the largest personal networks and a high reach to the rest of the network, while also being highly interconnected. These findings underscore the role of informal networks for facilitating change processes and can contribute to informing the practice of change agents and educational developers.

Change theories and the diffusion of innovation

Research in higher education has given a substantial amount of attention to understanding what makes change possible (Brankovic and Cantwell 2022). This is also partly due to the distinct context in which higher education exists as a social institution, which makes transferring knowledge from other domains (e.g. business) challenging (Kezar 2014). Some comprehensive studies on institutions’ change processes have been conducted (e.g. Kezar and Eckel 2002), which resulted in a multifaceted framework for understanding change (Kezar 2018).

One widely used theory for studying change and innovation is the Diffusion of Innovation theory, which refers to the process through which a certain innovation spreads within a system (e.g. organization, department) (Rogers 1983). The Diffusion of innovations theory (DoI) is employed in a variety of domains, including political sciences, public health, economy, and education (Stuart 2000). In education, some examples of its use include the implementation and the use of technology (Less 2003), awareness and adoption of education innovations (Borrego, Froyd, and Hall 2010; Jippes et al. 2013) and studying universities’ civic missions (Ćulum Ilić 2010). The key factor influencing the spread of innovation within a system is the disposition of the members of that system to accept, implement, and use a certain innovation (Ćulum Ilić 2010). Therefore, innovativeness is defined as a rate at which an individual adopts a new idea relatively earlier than other members of the system (Rogers 1983). He describes five categories of people related to their innovativeness, and their distribution within any system is idealized as normal distribution, as shown in Figure 1: Innovators (2.5%), Early adopters (13.5%), Early majority (34%), Late majority (34%), and Laggards (16%).

Figure 1. Adopter categorization based on innovativeness (Rogers 1983).

Accordingly, we will use the terms innovativeness or disposition towards change when describing teachers’ adoption rate relative to their peers. We establish their innovativeness based on their teaching habits and preferences as measured on a validated scale (see Data collection). Here we follow a similar approach as many other diffusion researchers who focus on the adoption as a tendency to use and implement a certain idea, rather than the implementation process itself (Rogers 1983). Those individuals who are more likely to adopt a new practice, idea, or an invention will therefore be considered more innovative. Because Rogers (1983) himself notes that the term laggard may carry some negative connotation and could be mistakenly used in a discriminatory way, we decided to use the term late adopter. Furthermore, even though we acknowledge there are many existing theories of change worth investigating, we opted for DoI because the use of a hypothetical normal frequency distribution curve to represent the distribution of adoption (Figure 1) is consistent with quantitative network-based diffusion models developed in other fields, such as agent-based social learning models in behavioral ecology (Franz and Nunn 2009; Reader 2004). Hence, a network perspective is complementary to DoI as a change theory (Kezar 2014).

One of the most important implications of Rogers’ theory is that it is not necessary to focus on convincing the entire system to adopt a change. In fact, Rogers’ (1983) analysis suggests that the widespread adoption of ideas depend on a relatively small percentage of members. In other words, diffusion of innovation happens gradually until it reaches a threshold, which in turn leads to the domino effect (Robertson 1967). That moment in which the innovation significantly increases its spread within a system is called the tipping point, and the domino effect is expected once the innovation is adopted by 10–20% of the total population of the system (Rogers 1983). This theory lends itself well to combination with social network analysis, particularly because some of the principal elements in a change process include interpersonal networks (Scott and Mcguire 2017). The role of interpersonal networks is also underscored in DoI as one of the four main factors that affect the adoption process is the social system – which includes both the formal structure of the individuals in the system and the informal structure of interpersonal networks linking all the individuals (Rogers 1983). In her review of educational change, Kezar (2014) points out that diffusion of innovation was linked to some studies of change and social networks, but mostly focusing on the use of technology and rarely on other processes.

Social network theory

Social network theory has been gaining increased attention in the field of educational research. Fundamentally, a social network consists of a finite set of actors (e.g. teachers) and the relations between them (e.g. informal conversations), where the relational information is the crucial and defining feature of a network and the focus of analysis (Wasserman and Faust 1994). It is one of the few theories within social sciences that is applicable to various levels of analysis from small groups to larger systems as organisations, nations, or even global systems (Kadushin 2012). Quantifying the relationships between actors makes the social structure visible and can be used to analyze relational and individual differences and behaviors within a social context (Clifton and Webster 2017). Compared to traditional social science research where the focal point of analysis lays in the individual and their attributes, social network analysis (while still able to provide such analyzes) offers an alternative perspective by providing a framework for studying the relations between these individuals (Wasserman and Faust 1994). This approach emphasizing relational data is central to the sociological tradition, despite being largely developed outside of the mainstream sociological research methods (Scott 2017).

As noted by Wasserman and Faust (1994), a network perspective has allowed researcher to take a different approach to study standard behavioral and social science research questions. Network analysis can be performed on multiple levels: individual or monadic level (e.g. various centrality measures of actors), relational or dyadic level (e.g. relationship patterns between individuals), and in case of our whole-network design, analysis can also be performed on the overall network level (e.g. density, transitivity, subgroups) (Clifton and Webster 2017; Wasserman and Faust 1994). Table 1 elaborates on all the network measures employed in this study.

Table 1. Network measures used in the study and the purpose of their use.

Network measure	Purpose
Network centralities	Identifying the most central or influential actors in a network
Degree centrality	Number of relationships an actor has in a network
Betweenness centrality	The extent to which an actor appears on the shortest path between other actors. Actors with high betweenness centrality are important brokers of information or resources.
Closeness centrality	The average distance between an actor and all other actors in a network. Actors with high closeness centrality are well-connected and have access to information from a large portion of the network.
Eigenvector centrality	Relative importance of an actor based on the centrality of the actors to which it is connected. Actors with high eigenvector centrality have many relationships to other central actors in the network.
Network density	The extent to which all actors in a network are connected
Network modularity	Measures the division of the network into smaller communities

Despite the growing significance of social network theory, research on higher education social networks was considered to be in its infancy (Van Waes et al. 2015), as literature focusing on networks in education tended to be more conceptual rather than empirical (de Lima 2010). In recent years there has been a noticeable increase of empirical social network studies in variety of topics, including professional and academic development programmes (e.g. Noben et al. 2022; Rienties et al. 2023; Rienties and Hosein 2015; Rienties and Kinchin 2014; Van Waes et al. 2015), collaborative work (Huang and Brown 2019), supervision and mentorship (Bäker, Muschallik, and Pull 2020; Jippes et al. 2013), students’ social support (Wilcox, Winn, and Fyvie-Gauld 2005) and international mobility (Macrander 2017).

As a result of this growing attention towards social networks, some significant empirical contributions have been made. Benbow and Lee (2019) used a mixed-method design to study the development of teaching-focused social networks and found differences between social network dimensions with regards to teaching experience, discipline, and institution type. Roxå and Mårtensson (2009) found that most university teachers rely on a limited number of significant others and that these significant networks are characterized by mutual trust, privacy, and intellectual intrigue. Additionally, a study that used qualitative social network analysis showed how the quality of instructors’ interactions varied depending on their developmental stage (novice, experienced non-expert, and expert), which provided implications for informing professional development of instructors (Van Waes et al. 2016).

However, an under-explored area where SNA could offer a substantial insight is in studying change and reform (Kezar 2014). A mixed-method exploratory case study found that social networks can facilitate or inhibit reform efforts and that careful analysis of social networks can contribute to our understanding of factors that support and constrain reforms (Daly and Finnigan 2010). Similarly, another study found that social networks may be as important as development course participations for clinical supervisors and suggested that social networks should be used to improve adoption of educational innovations (Jippes et al. 2013). As a result, social relationships among educators have received increased recognition as a resource for supporting educational innovation (Moolenaar and Sleegers 2010) as they provide a valuable framework for examining how teachers collaborate, discuss, exchange knowledge, information, and other resources which can often become levers for innovative practice (Van Waes et al. 2016). This approach challenges widespread beliefs that formal educational structures have a dominant impact on individuals’ choices (i.e. top-down) and suggest that informal networks also play a role in influencing their decisions to engage and accept change or reform (Kezar 2018). In other words, social networks can be leveraged to accelerate and enhance change processes as well as improve diffusion and dissemination of innovation (Valente 2012). Despite these significant findings, the relationship between higher education teachers’ disposition towards change and their network characteristics remains largely unknown, and we are in a unique position to contribute to the understanding of that relationship on account of capturing complete departmental networks within which these teachers act.

Methodology

Participants and data collection

The setting for our research is a university in Norway. We use data collected from a department within STEM disciplines, and the target population for this study included department teaching staff who had been involved in teaching duties within the previous 6 months of the distribution of the questionnaire. While some SNA studies take a sample from a larger population and examine the actors’ individual networks, this study aims at capturing the entire population within its network boundaries (i.e. whole-network or sociocentric design). This type of network can be generated only in situations where the boundaries of the network are clearly defined. The main benefit of this approach lies in its capacity for a more comprehensive level of analysis, including measuring correlation between different networks, detecting subgroups, analyzing potentially more relevant centrality measures (e.g. eigenvector centrality), which we would otherwise fail to capture.

While there are certain limitations associated with this approach in terms of its capacity for generalization, collecting data on an entire departmental network is a crucial step in understanding the network structure with regards to innovativeness. Some authors suggest that using a whole-network approach is a preferable strategy as it provides a more complete picture of the network and supports the validity of results (Scott 2017). This argument aligns well with studies concerning change, as researchers are emphasizing the need to shift from the broad generalizations towards focusing on more detailed methods (Kezar and Eckel 2002). A whole-network approach will also allow us to investigate the characteristics of smaller communities in the network as well as explore the teachers’ positions in the network relative to their innovativeness and other independent variables.

The questionnaire was sent out to 72 teaching staff working at the same department within STEM disciplines at a Norwegian university with academic positions ranging from PhD candidates to Full professors. From that, 42 teaching staff  (59.16%) have completed and returned the survey, and these 42 participants have nominated 71 teachers (98.61%) for the teaching network and 70 actors (97.22%) for the research network, which gives us an almost complete image of the teaching and research networks at the department. This is otherwise known as the reconstruction method (Stork and Richards 1992) in non-directional networks used to infer non-respondents’ personal networks based on other respondents. In other words, a population of 72 teachers yields 6,184 possible relationships (ties). Our study was able to map 4,284 (82.64%) of these relationships. This includes 3,024 connections identified directly from the raw data and an additional 1,260 connections uncovered through our reconstruction method. However, our approach cannot determine connections between teachers who did not respond to our survey, accounting for 900 potential relationships.

As a result, our investigation of RQ1 is applicable to 71 teachers (98.61% of the total population). In contrast, our analysis of RQ2 is constrained to the 42 participants who completed the DoI survey. We conducted a non-response bias analysis to compare respondents and non-respondents. The results indicated no statistically significant differences between these two groups, suggesting that the non-response did not introduce a notable bias into our study findings.

We utilized a multi-method approach for analyzing teachers’ informal teaching networks and their innovativeness. For descriptive exploration and visual representations of the networks displayed in this study, we used the software platform Gephi. For statistical analysis of the social network data, we used UCINET software package (Borgatti, Everett, and Freeman 2002). To analyze network data, we computed a Quadratic Assignment Procedure (QAP) based on 10,000 permutations. QAP is used to correlate two matrices by comparing the observed correlation to correlations between thousands of pairs of matrices (Borgatti, Everett, and Johnson 2013). Finally, to analyze the DoI survey data, we utilized the SPSS statistical software platform in conjunction with Python (Pandas, Numpy, and Scipy packages) to integrate network results with survey data and conduct descriptive and inferential statistics.

Survey instrument

Considering the proposed theoretical framework and the embedded research questions, this study focuses on collecting network data from higher education teachers and comparing the network results to a quantitative survey exploring their disposition to implement changes in their teaching. To do so, we developed and distributed an online questionnaire that consists of two types of data.

Sociometric network questions. The purpose of the network survey is to capture the department’s teaching and research networks. Participants were provided a list (i.e. roster) of all teaching staff at the department. It is recommended to always use a roster if the network structure permits it (Stork and Richards 1992), as it allows for identification of less-salient, weak connections which are otherwise easier to forget (Clifton and Webster 2017). The network survey asked teacher to indicate all colleagues they had talked to in the last 6 months, with the first part asking about teaching-related topics, and the second part asking about research-related topics. This resulted in two separate networks – one for teaching conversations and one for research conversations. Participants could choose only from the list and could not add any other potential teachers outside of the department. Hence, our sampling strategy is known as the nominalist approach (i.e. researcher-defined networks) where all individuals who match a certain criterion (i.e. members of a department) are part of the population (Borgatti, Everett, and Johnson 2013). This does not imply that relationships outside of the department do not exist or are unimportant, but it merely specifies that the study is about inter-departmental networks. Similarly, to Van Waes et al. (2018), we have clarified that administrative discussions about teaching and research should not be considered when specifying their informal networks. We have opted for collecting valued data where the frequency of every potential relationship is assessed on a Likert scale ranging from 1 – talked once in the last 6 months to 5 – talked multiple times a week. In order to be able to measure the density of the two networks, there was no limit to the number of colleagues a teacher could nominate from the roster.

Disposition for Teaching Change scale. This part of the questionnaire is based on the Diffusion of Innovations theory (Rogers 1983), and the items were taken from Ćulum Ilić (2010) who designed and tested the scale aiming to investigate higher education teachers’ disposition towards introducing pedagogical innovations. Participants were given 17 statements and asked to indicate the extent to which the statements refer to their practices on a scale from 1-completely disagree to 5-completely agree, and Cronbach’s alpha (α = .84) demonstrates that the survey is reliable. Finally, once we confirmed that the survey results followed normal distribution, we categorized the participants into 4 categories of DoI based on their responses: participants with responses of +1 + 3 SD were categorized as Innovators and Early Adopters, from 0 to (+1) SD Early Majority, from 0 to (−1) SD Late Majority, and from −1 to (−3) Late adopters. Compared to the original classification into 5 categories, we decided to merge Innovators (2.5%) and Early adopters (13.5%) into one category due to the smaller sample size and for the simplicity of further statistical inquiry (Figure 2).

Figure 2. Distribution of the results in the DOI survey and the subsequent categorization.

Results

Network characteristics of teachers’ informal teaching networks (RQ1)

The first set of results is concerned solely with the network survey and the associated research question. To answer that question, we have generated two separate networks – one for teaching conversations and one for research conversations – and the results highlight comparable network structures. As the descriptive results show comparable graph densities of the teaching network (D = 0.136) and the research network (D = 0.121), we wanted to explore the extent to which these two networks are connected. The results demonstrate a strong, statistically significant Pearson correlation between the two networks (r = .577, p < .001). In other words, if two teachers are engaging in conversation about research-related topics, they are highly likely to engage in conversation about teaching-related topics and vice versa.

When analyzing network data, it is also important to consider which factors could influence the formation of relationships. We used the UCINET’s cross-product matrix operation that counted how many conversations partners each pair of teachers have in common. Afterwards, we conducted a QAP correlation and found a statistically significant moderate correlation (r = 0.410, p < .001) between the number of conversation partners teachers have in common and the extent to which they talk to each other. This result reflects the tendency to interact more with those people with whom you have other conversation partners in common. To explore what made those relationships emerge and maintain themselves, we ran a community detection algorithm (Blondel et al. 2008) that detects subgroups in the network, which are solely comprised of ties between individuals, devoid of any demographic data. The algorithm revealed 4 main subgroups within the teaching network. We compared the subgroups with teachers’ demographic characteristics to explore whether there are any shared characteristics that could explain the formation of these subgroups. We found that in 38 out of 42 participants, there is an overlap between the algorithm-based subgroup and the departmental research groups the teachers belong to (Figure 3). That means that because there are more informal teaching conversations within a smaller disciplinary group than across it, they serve as a hub for the group of people involved in teaching.

Figure 3. Network visualization of the overlap between algorithm-detected subgroups departmental research groups. The four research groups are labeled as A, B, C, D, and the four detected communities are also labeled as A, B, C, D. In cases where there is an overlap between the two, the nodes are gray, and in the four cases where there is no overlap, the nodes are red.

Furthermore, the network survey was designed so that it can capture both strong and weak ties. We use frequency of conversations as a measurement for tie strength. Teachers who engaged in conversation once in six months or a few times in six months (values 1 or 2 on the Likert scale) have a weak tie, and those who engage in conversation monthly, weekly, or several times per week (values 3, 4, or 5 on the Likert scale) have a strong tie. We analyzed to what extent the 338 teaching relationships are a result of weak and strong ties. Out of the 338 ties, only 62 of them (∼18%) are frequent teaching conversations that occur monthly or more often (Figure 4). In other words, most informal conversations (∼82%) are weak ties that had happened only once or a few times in the last 6 months.

Figure 4. Comparison of the complete teaching network (338 ties) with the teaching network consisting of strong ties (62 ties).

Relationship between higher education teachers’ social network structure and their disposition towards implementing changes in teaching (RQ2)

To address the second research question, we compared the network data with the results from the DoI survey, to explore if there is a connection between various measures of teachers’ network centralities and their innovativeness. We computed Pearson correlation coefficient and found that teachers’ disposition to implement changes is moderately and positively correlated with eigenvector centrality (r(40) = .37, p < .05), closeness centrality (r(40) = .33, p < .05), and degree centrality (r(40) = .32, p < .05). In other words, the result shows that those who tend to have more informal conversations with peers also tend to be the most innovative. Figure 5 shows a monotonic relationship for three upper categories excluding late adopters, with innovators and early adopters (M = 0.72) having the highest eigenvector centrality. However, the analysis of strong ties shows a monotonic relationship for all categories of innovativeness with regards to centrality.

Figure 5. Comparison of the influence on the network (i.e. eigenvector centrality) of different categories of innovativeness for both the full network and frequent conversations (i.e. strong ties).

Furthermore, after noting that those teachers who have the highest disposition towards introducing changes also tend to be the most central, we decided to explore the extent of their relationships within the entire network. By extracting all personal relationships (i.e. ego-networks) the Innovators and EA category, we found that they engage in conversation to a total of 43 other teachers excluding themselves (∼66% of the network) and they are responsible for a total of 113 ties (∼33% of total ties) in the network (Figure 6). In other words, 16% of the most innovative teachers have conversations with ∼66% of the network and are responsible for ∼33% of all informal teaching conversations that occur in a period of 6 months.

Figure 6. Combined personal networks of most innovative teachers. Black nodes indicate Innovators and Early adopters, gray nodes indicate other teachers.

Finally, apart from noting that most innovative teaching staff tend to have a high reach to the rest of the network, we also observed a difference when analyzing the within-category conversations. We define within-category conversations as those relationships that occur only between teachers belonging to the same innovativeness category. Figure 7 shows the difference between within-category network structures and reveals that Innovators and EA have the highest network density (D = 0,800) of all categories. This result indicates that Innovators and EA do not only engage in conversation more often across the whole network, but they also engage in conversation with each other more often.

Figure 7. Graph densities of Innovators and Early adopters (1st), Early majority (2nd), Late majority (3rd), Late adopters (4th).

Discussion

Our study aimed at investigating the role and characteristics of teachers’ social interactions with regards to their disposition towards introducing changes. The results can be summarized into two main categories: overall department network characteristics and characteristics of innovative teachers. Firstly, we found that there is a strong interplay between teaching and research networks and that formal organizational sub-units serve the role of hubs for teaching conversations. Secondly, we found that the most innovative teachers tend to have the largest personal networks and a high reach to the rest of the network, while also being highly interconnected.

We can draw several implications from these results that might provide recommendations for change agents engaged in promoting change in higher education. The role of the teaching-research nexus is a topic of debate in the higher education literature (Annala et al. 2022; Barnett 1992; Coate, Barnett, and Williams 2001; Leisyte, Enders, and de Boer 2009; Skvoretz et al. 2023; Verburgh, Elen, and Lindblom-Ylänne 2007). This study provides empirical evidence that teachers tend to interact with same colleagues about both teaching and research. Although our aim was not to determine whether there is a positive or negative relationship between teaching and research activities, we can still conclude that teachers tend to discuss both teaching and research-related topics with the same colleagues. Therefore, given the perceived interconnectedness of teaching and research, encompassing both positive and negative relationships (Coate, Barnett, and Williams 2001), it is necessary to establish explicit strategies aimed at facilitating this potential synergy. Coate, Barnett, and Williams (2001, 162) claim that what is missing is ‘ … a managerial strategy that promotes the intellectual perception of teaching and research as integrated.’ Therefore, change could potentially also be framed more holistically, involving both teaching and research-related activities (Annala et al. 2022).

The fact that teachers engage in both conversations with the same colleagues should encourage new strategies that will utilize the overlapping networks to facilitate successful positive relationships between the two. This argument is also echoed by some educational developers who acknowledge the teaching-research nexus as an important dimension for developing positive attitudes towards teaching (Miočić, Brajdić Vuković, and Ledić 2020). Not only do we find a strong relationship between teaching and research networks, but our results also show that teachers tend to build smaller communities around their respective research groups where they engage in more frequent conversations about teaching, which has several important implications. Primarily, our findings lend strong support to the idea of the role of territory at the departmental level (Anakin et al. 2018) and that these territories can play a vital role in inhibiting or facilitating change (Trowler, Saunders, and Bamber 2012). Our study expands on Quardokus and Henderson (2015) who noted that attention should be given to the departmental subgroups as an important avenue for change agents’ activity, by suggesting that change agents should focus their effort on the smaller organizational sub-units (e.g. research groups) to promote change. This approach may be more productive than larger scale attempts as these smaller communities are based on trust, a common developmental agenda, and a shared responsibility (Roxå and Mårtensson 2015) because teachers generally discuss teaching with a core personal network (Benbow, Lee, and Hora 2021). The importance of core personal networks is further supported by another study showing that local networks based on collegial understanding and sharing were vital in supporting agency (Annala et al. 2023).

By conducting network analysis on an entire department, we were able to pinpoint how these communities revolve around the respective research groups. These results can inform change agents by suggesting that the nucleus of activity could be placed within the research groups, where teachers engage in informal conversations more often and where the associated risk of a change process is collective and not individual. Moreover, our results also have implications for guiding teacher training and academic development, which often raises questions of what organizational level should be the locus of training (Roxå and Mårtensson 2012). Even though teachers’ personal networks tend to revolve around shared educational beliefs (Poole, Iqbal, and Verwoord 2019), analysis of both frequent and infrequent conversations indicates that these strong ties grounded in shared beliefs propagate to larger formal departmental sub-units when infrequent conversations are considered.

Our findings also shed light on some distinctive network characteristics of the diffusion groups. Here we shift our focus towards the most engaged and innovative teachers, particularly because Rogers’ theory (1983) suggests that these 10–20% are considered a critical mass after which the adoption rate significantly increases. Our results can be summarized through three distinct network characteristics of innovators: network size, network reach, and interconnectedness. When looking at the category of Innovators and Early Adopters (who comprise 16% of the system), it is evident that they on average have the largest personal networks and have the greatest access to information. This result is similar to a study from Knaub, Henderson, and Fisher (2018) who found that those who were identified as leaders in change initiatives also tend to have more ties on average compared to non-leaders. This pattern becomes even more prominent when comparing the full network to only strong ties (Figure 4). The main distinction between strong and weak ties lies in their respective purpose: weak ties facilitate access to information and play an important in the spread of information, whereas strong ties are more likely to have stronger influence over decisions (Granovetter 1983). Our finding that there is a monotonic relationship between different categories of innovativeness for eigenvector centrality in strong ties reinforces the notion that the most innovative teachers, who are the most central and influential actors with respect to teaching innovations, should be a focal point for change agents seeking to effect meaningful transformations.

Furthermore, several scholars have discussed similar bottom-up processes that can motivate successful change. For example, Quardokus Fisher and Henderson (2018) discuss department-level change processes and Kezar (2018) discusses several different change theories as analytic tools for change agents to consider. Our study also highlights the potential benefit of pinpointing the most innovative members of a system because not only do they talk with many other individuals, but they also develop a web of relationships that envelops much of the network. In addition, the high interconnectedness among the most innovative actors in our study supports Rogers (1983, 248) observation that ‘communication patterns and friendships among a clique of innovators are common, even though the geographical distance between the innovators may be considerable.’ In other words, our study confirms that innovators tend to be the most active and integrated members of the system in which they participate, which is something previously hypothesized by Ćulum Ilić (2010).

More importantly, the tendency for the most innovative teachers to be central in their respective subgroups provides an important heuristic for change agents when considering how to increase the potential success of implementing change processes across the department level. The combination of high network centrality and reach of innovators also seems to favor a cross-scale approach to departmental-level change, involving both top-down and bottom-up change processes. Specifically, innovators may play a pivotal role in driving change by influencing other individuals within the network due to their extensive reach, which holds the potential to catalyze changes at the meso-scale (organizational sub-units) and macro-scale (department level or beyond). Therefore, the discovered role of innovators in informal networks supports the rationale for directing organizational and resource support towards them, rather than expending efforts on persuading the most resistant individuals about the advantages of change. Particularly when viewed through the lens of educational change as a time-intensive cultural transformation, the amount of support teachers require should not be underestimated (Keesing-Styles, Nash, and Ayres 2014). A similar study by Jippes et al. (2013) concluded that development efforts should capitalize on faculty members’ networks to facilitate implementation of educational innovations in medicine. We further support that suggestion by providing the specific network characteristics of the most innovative teachers, who tend to have greater centrality, interconnectedness, and reach.

Limitations

It is important to acknowledge certain limitations of this study. The first limitation concerns the generalizability of findings. Focusing on a whole-network approach in conducting social network analysis means that we are not analyzing a sample from a large population, but rather a smaller population itself (Borgatti, Everett, and Johnson 2013), which in turn makes any wider generalizations challenging. However, we have opted for this approach because we wanted to analyze not only the relationship between personal networks and individual characteristics, but also to investigate the interplay between these individual members and the entire network. Furthermore, there is a growing understanding that change processes are better studied through more detailed and in-depth methodological approaches that go beyond broad generalizations (Kezar and Eckel 2002). These reasons made a whole-network approach a priority at the cost of generalizability of the results, while still offering important implications for further research.

Another limitation is that this is a cross-sectional quantitative study, which means that there are certain constraints in our ability to infer temporal or causal relationships. For example, we are unable to determine whether research network causes the teaching network or vice versa, nor can we determine other implications related to the spread of information that could be inferred through longitudinal studies. As a quantitative study, this study is restricted in its capacity to obtain deeper and more nuanced insight into various adoption categories and their perspectives. Our study does not encompass the broader spectrum of cross-organizational and interdepartmental networks.

Future research could explore these wider connections, examining the recognized influence of external ties as highlighted in the concepts of invisible college (Crane 1972), boundary crossing (Rienties and Héliot 2018), or significant networks (Roxå and Mårtensson 2009), to provide a more comprehensive understanding of academic collaborations. Furthermore, future studies should aim to expand this approach by collecting qualitative data and using existing models for mixed-methods SNA to understand the nature of these networks whilst addressing the single-method limitations. Finally, it would also be beneficial to collect data longitudinally to track the evolution of networks and investigate which network characteristics are more prone to changes.

Conclusion

In summary, our study contributes to a growing body of research on how to facilitate successful change initiatives in higher education. By providing detailed insight into the relationship between the teaching networks of higher education teachers and their disposition to implement changes, our results provide implications for both policy and practice. On the one hand, quantitative insight into the strong interrelation between research and teaching might be useful for enhancing the effectiveness of academic development programs by leveraging the potential synergy between these two spheres of academic life. On the other hand, the strong interdependency between communities of teaching and research might also exacerbate disciplinary fragmentation and compartmentalization of domains of knowledge. Our study also provides robust quantitative evidence of network characteristics of the most innovative teachers, revealing their centrality, extensive reach to the rest of the department, and high interconnectivity. These findings can guide organizational efforts to support change processes and foster cultural transformation within higher education departments by focusing on these influential actors. Our study confirms the potential of social network analysis as a powerful methodology in (higher) education research that enables an exploration of complex relationships, network structures, and patterns. Notably, the emphasis on complete networks based on roster and valued data has yielded significant insights into the patterns of frequent and infrequent informal conversations that may have otherwise been overlooked. This approach can not only give researchers a detailed picture of teaching interactions, but coupled with traditional research approaches (e.g. surveys, interviews), can help reveal individual and social mechanisms that shape the dynamics of change in higher education.

Ethics approval

Approval was gained from the Data Protection Services from the Norwegian Agency for Shared Services in Education and Research (Sikt) who have carried out the assessment of the processing of personal data and confirmed that the processing will comply with data protection legislation.

Consent

All participants have received an informed consent form clarifying the project's purpose, the use of personal data, and their rights. Before participating in the study, the participants signed the form giving consent for their personal data to be stored in an anonymized version for the duration of the project.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The datasets analyzed during the current study are not publicly available as part of data protection legislation, but the modified version based on anonymized and deidentified data is available from the corresponding author on reasonable request.

Additional information

Funding

This work was supported by the iEarth Centre for Excellence in Education (SFU), funded by the Norwegian Directorate for Higher Education and Skills.